It is not traditional phase noise, and not a normal Nyquist filtering problem, but, yes, presence of significant second harmonic energy in your DDS output will shift/dither your squaring input.
It is hard, with practical filters and filter components to get much better coverage than one-half octave per low-pass filter. It will be set by how much second harmonic content you can tolerate. That is, how far down the low-pass filter slope the second harmonic needs to be to get the performance you want. So using the one-half octave per filter "rule of thumb" then you might need three or more low pass filters, depending on the actual frequency span output you are dealing with. --- Graham On Thu, Jul 28, 2016 at 6:39 PM, life speed via time-nuts < [email protected]> wrote: > Hi, > > Been a while since I visited, I recall there are many, well, time nuts > here. I am trying to track down a source of phase noise in a frequency > synthesizer design. One part of the frequency reference upconverts a DDS > and then divides it down again using a digital divider - standard technique > for DDS angle modulation spurious reduction. > > The DDS tunes over more than an octave, so obviously the single low pass > filter isn't going to cut it. I am noticing up to 3 dB phase noise > degradation at the output of the divider as the DDS frequency decreases and > the 2nd harmonic is in-band to the LPF. I suspect this is disturbing the > threshold crossing in the high speed digital logic divider, as described in > "The Effect of Harmonic Distortion on Phase errors in Frequency > Distribution and Synthesis" by F.L. Walls et al at NIST. > > What do you think? I should probably put in a switched filter to get my 3 > dB back ;) > > - Lifespeed > _______________________________________________ > time-nuts mailing list -- [email protected] > To unsubscribe, go to > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. > _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
